Meta’s BlenderBot is more like a ‘Blunder-Bot’
Meta has come under fire for not doing enough to stop the spread of misinformation and hate speech on its platforms. An ex-employee last year, Frances Haugen, claimed the business prioritised profits over online security. Facebook, Facebook Messenger, Instagram, and WhatsApp, four of the biggest social media platforms and messaging services in the world, are all owned by the business. The algorithm in BlenderBot 3 researches the web to inform its conclusions. It’s possible that the algorithm’s opinions on Mr Zuckerberg have “learned” from other people’s ideas that it has studied.
What does the BlenderBot actually do?
AI software such as chatbots is used to simulate human conversations using text and audio. They are frequently employed in customer service or voice assistants. Companies are working to enhance chatbots’ capabilities as users spend more time interacting with them. The artificial intelligence-powered bot, known as BlenderBot 3, is intended to enhance its safety and conversational abilities by speaking with people. Meta recently publicly launched the chatbot as part of an AI research initiative. On this open website, US people can have conversations with Meta’s new chatbot on almost any subject. The AI constructs its messages using recollections of past chats as well as internet searches.
Blenderbot has been doing Blunders
However, the new AI chatbot for Meta has been spewing out unflattering information about Meta and its CEO Mark Zuckerberg. When questioned about Zuckerberg, the bot responded as follows: “He is a good businessman, but his business practices are not always ethical. It is funny that he has all this money and still wears the same clothes!” People are being invited by Meta to speak with BlenderBot and provide developers with their valuable feedback. Microsoft, the parent firm of LinkedIn, attempted to train its “Tay” bot using Twitter users in 2016, but the effort immediately backfired when the bot made offensive racial and political statements.
According to a blog post about the new chatbot by Meta, researchers analysed data that is generally gathered through experiments where participants interact with bots in a controlled setting. But that data collection doesn’t adequately represent variety around the world, so researchers are turning to the public for assistance. Although researchers have shown that the chatbot gets safer the more it learns from interacting with people, Meta admitted that safety is still an issue.
“A live demo is not without challenges, however,” the blog post by Meta said. “It is difficult for a bot to keep everyone engaged while talking about arbitrary topics and to ensure that it never uses offensive or toxic language.”